I scraped 5,000+ NDE/OBE accounts from public research databases, extracted structured data, generated embeddings, and used UMAP to project them into 3D space.

The result is an interactive "consciousness map" where similar experiences cluster together.

Tech stack: Next.js, Supabase (pgvector), Vercel, OpenAI API, Three.js.

Some interesting patterns: - "Void" experiences cluster separately from "light" experiences - High Greyson scores tend to include more entity encounters - Cardiac arrest NDEs cluster differently than drowning NDEs

Happy to answer questions about the AI pipeline or the data.